Overhaul warning alerts

charts/seed-bootstrap/templates/alertmanager/_config.tpl

@@ -30,7 +30,6 @@ route:
   # email only for critical and blocker
   - match_re:
       visibility: ^(all|operator)$
-      severity: ^(blocker|critical|info)$
     receiver: email-kubernetes-ops
 
 

charts/seed-monitoring/charts/alertmanager/templates/_config.tpl

@@ -27,7 +27,6 @@ route:
   # email only for critical and blocker
   - match_re:
       visibility: ^(all|owner)$
-      severity: ^(critical|blocker|info)$
     receiver: email-kubernetes-ops
 
 inhibit_rules:

charts/seed-monitoring/charts/prometheus/rules-tests/apiserver-connectivity-check.rules.test.yaml

@@ -6,6 +6,7 @@ evaluation_interval: 30s
 tests:
 - interval: 30s
   input_series:
+  # ApiServerUnreachableViaKubernetesService
   - series: 'probe_success{job="blackbox-exporter-k8s-service-check"}'
     values: '0+0x30'
   alert_rule_test:

charts/seed-monitoring/charts/prometheus/rules-tests/cloud-controller-manager.rules.test.yaml

@@ -6,6 +6,7 @@ evaluation_interval: 30s
 tests:
 - interval: 30s
   input_series:
+  # CloudControllerManagerDown
   - series: 'up{job="cloud-controller-manager"}'
     values: '0+0x30'
   alert_rule_test:

charts/seed-monitoring/charts/prometheus/rules-tests/coredns.rules.test.yaml

@@ -6,6 +6,7 @@ evaluation_interval: 30s
 tests:
 - interval: 30s
   input_series:
+  # CoreDNSDown
   - series: 'up{job="coredns"}'
     values: '0+0x40'
   alert_rule_test:

charts/seed-monitoring/charts/prometheus/rules-tests/kube-apiserver.rules.test.yaml

@@ -6,14 +6,22 @@ evaluation_interval: 30s
 tests:
 - interval: 30s
   input_series:
+  # ApiServerNotReachable
   - series: 'probe_success{instance="api.example.com/healthz", job="blackbox-apiserver"}'
     values: '0+0x10'
+  # KubeApiserverDown
   - series: 'up{job="kube-apiserver"}'
     values: '0+0x10'
+  # KubeApiServerTooManyOpenFileDescriptors
   - series: 'process_open_fds{job="kube-apiserver", instance="instance"}'
     values: '81+0x60'
   - series: 'process_max_fds{job="kube-apiserver", instance="instance"}'
     values: '100+0x60'
+  # KubeApiServerLatency
+  - series: 'apiserver_request_latencies_bucket{verb="POST", le="5000000"}'
+    values: '100+0x60'
+  - series: 'apiserver_request_latencies_bucket{verb="POST", le="+Inf"}'
+    values: '100+0x60'
   alert_rule_test:
   - eval_time: 5m
     alertname: ApiServerNotReachable
@@ -48,6 +56,7 @@ tests:
         type: seed
         job: kube-apiserver
         instance: instance
+        visibility: owner
       exp_annotations:
         description: 'The API server (instance) is using 81% of the available file/socket descriptors.'
         summary: 'The API server has too many open file descriptors'
@@ -61,4 +70,16 @@ tests:
       exp_annotations:
         description: 'The API server (instance) is using 81% of the available file/socket descriptors.'
         summary: 'The API server has too many open file descriptors'
+  - eval_time: 30m
+    alertname: KubeApiServerLatency
+    exp_alerts:
+    - exp_labels:
+        service: kube-apiserver
+        severity: warning
+        type: seed
+        visibility: owner
+        verb: POST
+      exp_annotations:
+        description: Kube API server latency for verb POST is high. This could be because the shoot workers and the control plane are in different regions. 99th percentile of request latency is greater than 3 second.
+        summary: Kubernetes API server latency is high
 

charts/seed-monitoring/charts/prometheus/rules-tests/kube-controller-manager.rules.test.yaml

@@ -6,6 +6,7 @@ evaluation_interval: 30s
 tests:
 - interval: 30s
   input_series:
+  # KubeControllerManagerDown
   - series: 'up{job="kube-controller-manager"}'
     values: '0+0x30'
   alert_rule_test:

charts/seed-monitoring/charts/prometheus/rules-tests/kube-etcd3.rules.test.yaml

@@ -6,14 +6,21 @@ evaluation_interval: 30s
 tests:
 - interval: 30s
   input_series:
+  # KubeEtcdMainDown
   - series: 'up{job="kube-etcd3",role="main"}'
     values: '0+0x20'
+  # KubeEtcdEventsDown
   - series: 'up{job="kube-etcd3",role="events"}'
     values: '0+0x30'
+  # KubeEtcd3MainNoLeader
   - series: 'etcd_server_has_leader{job="kube-etcd3",role="main"}'
     values: '0+0x20'
+  # KubeEtcd3EventsNoLeader
   - series: 'etcd_server_has_leader{job="kube-etcd3",role="events"}'
     values: '0+0x30'
+  # KubeEtcd3HighNumberOfFailedProposals
+  - series: 'etcd_server_proposals_failed_total{job="kube-etcd3", pod="etcd"}'
+    values: '0+0x15 1+0x15 2+0x15 3+0x15 4+0x15 5+0x15 6+0x15 7+0x15'
   alert_rule_test:
   - eval_time: 5m
     alertname: KubeEtcdMainDown
@@ -58,4 +65,18 @@ tests:
         visibility: operator
       exp_annotations:
         description: Etcd3 events has no leader. No communication with etcd events possible. New cluster events cannot be collected. Events can only be read.
-        summary: Etcd3 events has no leader.
+        summary: Etcd3 events has no leader.
+  - eval_time: 1h
+    alertname: KubeEtcd3HighNumberOfFailedProposals
+    exp_alerts:
+    - exp_labels:
+        service: etcd
+        severity: warning
+        type: seed
+        visibility: operator
+        pod: etcd
+        job: kube-etcd3
+      exp_annotations:
+        description: Etcd3 pod etcd has seen 7 proposal failures
+          within the last hour.
+        summary: High number of failed etcd proposals

charts/seed-monitoring/charts/prometheus/rules-tests/kube-kubelet.rules.test.yaml

@@ -6,16 +6,24 @@ evaluation_interval: 30s
 tests:
 - interval: 30s
   input_series:
+  # KubeKubeletNodeDown
   - series: 'up{job="kube-kubelet", type="shoot"}'
     values: '0+0x60'
+  # KubeletTooManyOpenFileDescriptorsShoot
   - series: 'process_open_fds{kubernetes_io_hostname="hostname", job="kube-kubelet"}'
     values: '81+0x60'
   - series: 'process_max_fds{kubernetes_io_hostname="hostname", job="kube-kubelet"}'
     values: '100+0x60'
+  # KubeletTooManyOpenFileDescriptorsSeed
+  - series: 'process_open_fds{kubernetes_io_hostname="hostname", job="kube-kubelet-seed"}'
+    values: '81+0x60'
+  - series: 'process_max_fds{kubernetes_io_hostname="hostname", job="kube-kubelet-seed"}'
+    values: '100+0x60'
+  # KubePersistentVolumeUsageCritical KubePersistentVolumeFullInFourDays
   - series: 'kubelet_volume_stats_available_bytes{job="kube-kubelet", type="seed", persistentvolumeclaim="pvc1"}'
-    values: '2+0x60'
+    values: '0+0x240'
   - series: 'kubelet_volume_stats_capacity_bytes{job="kube-kubelet", type="seed", persistentvolumeclaim="pvc1"}'
-    values: '100+0x60'
+    values: '100+0x240'
   alert_rule_test:
   - eval_time: 30m
     alertname: KubeletTooManyOpenFileDescriptorsShoot
@@ -41,15 +49,42 @@ tests:
         description: 'Shoot-kubelet (hostname) is using 81% of the available file/socket descriptors. Kubelet could be under heavy load.'
         summary: 'Shoot-kubelet has too many open file descriptors.'
   - eval_time: 30m
+    alertname: KubeletTooManyOpenFileDescriptorsSeed
+    exp_alerts:
+    - exp_labels:
+        job: kube-kubelet-seed
+        service: kube-kubelet-seed
+        severity: critical
+        visibility: operator
+        type: seed
+        kubernetes_io_hostname: hostname
+      exp_annotations:
+        description: 'Seed-kubelet (hostname) is using 81% of the available file/socket descriptors. Kubelet could be under heavy load.'
+        summary: 'Seed-kubelet has too many open file descriptors.'
+  - eval_time: 1h
     alertname: KubePersistentVolumeUsageCritical
     exp_alerts:
     - exp_labels:
         job: kube-kubelet
+        service: kube-kubelet-seed
         severity: critical
         type: seed
         visibility: operator
         persistentvolumeclaim: pvc1
       exp_annotations:
-        description: The PersistentVolume claimed by pvc1 is only 2.00% free.
+        description: The PersistentVolume claimed by pvc1 is only 0.00% free.
         summary: PersistentVolume almost full.
+  - eval_time: 2h
+    alertname: KubePersistentVolumeFullInFourDays
+    exp_alerts:
+    - exp_labels:
+        job: kube-kubelet
+        service: kube-kubelet-seed
+        severity: warning
+        type: seed
+        visibility: operator
+        persistentvolumeclaim: pvc1
+      exp_annotations:
+        description: Based on recent sampling, the PersistentVolume claimed by pvc1 is expected to fill up within four days. Currently 0.00% is available.
+        summary: PersistentVolume will be full in four days.
 

charts/seed-monitoring/charts/prometheus/rules-tests/kube-pods.rules.test.yaml

@@ -0,0 +1,77 @@
+rule_files:
+- ../rules/kube-pods.rules.yaml
+
+evaluation_interval: 30s
+
+tests:
+- interval: 30s
+  input_series:
+  # KubePodPendingShoot
+  - series: 'kube_pod_status_phase{phase="Pending", type="shoot", pod="pendingPod"}'
+    values: '1+0x120'
+  - series: 'kube_pod_labels{label_origin="gardener", pod="pendingPod"}'
+    values: '1+0x120'
+  # KubePodPendingControlPlane
+  - series: 'kube_pod_status_phase{phase="Pending", type="seed", pod="pendingPod"}'
+    values: '1+0x60'
+  # KubePodNotReadyShoot
+  - series: 'kube_pod_status_ready{condition="true", type="shoot", pod="notReadyPod"}'
+    values: '0+0x120'
+  - series: 'kube_pod_labels{label_origin="gardener", pod="notReadyPod"}'
+    values: '1+0x120'
+  # KubePodNotReadyControlPlane
+  - series: 'kube_pod_status_ready{condition="true", type="seed", pod="cpPodNotReady"}'
+    values: '0+0x60'
+  alert_rule_test:
+  - eval_time: 1h
+    alertname: KubePodPendingShoot
+    exp_alerts:
+    - exp_labels:
+        service: kube-kubelet
+        severity: warning
+        visibility: owner
+        type: shoot
+        pod: pendingPod
+        phase: Pending
+      exp_annotations:
+        description: Pod pendingPod is stuck in "Pending" state for more than 1 hour.
+        summary: Shoot pod stuck in "Pending" state
+  - eval_time: 30m
+    alertname: KubePodPendingControlPlane
+    exp_alerts:
+    - exp_labels:
+        service: kube-kubelet
+        severity: warning
+        visibility: operator
+        type: seed
+        pod: pendingPod
+        phase: Pending
+      exp_annotations:
+        description: Pod pendingPod is stuck in "Pending" state for more than 30 minutes.
+        summary: Control plane pod stuck in "Pending" state
+  - eval_time: 1h
+    alertname: KubePodNotReadyShoot
+    exp_alerts:
+    - exp_labels:
+        service: kube-kubelet
+        severity: warning
+        visibility: owner
+        type: shoot
+        pod: notReadyPod
+        condition: true
+      exp_annotations:
+        description: Pod notReadyPod is not ready for more than 1 hour.
+        summary: Shoot pod is in a not ready state
+  - eval_time: 30m
+    alertname: KubePodNotReadyControlPlane
+    exp_alerts:
+    - exp_labels:
+        service: kube-kubelet
+        severity: warning
+        visibility: operator
+        type: seed
+        pod: cpPodNotReady
+        condition: true
+      exp_annotations:
+        description: Pod cpPodNotReady is not ready for more than 30 minutes.
+        summary: Control plane pod is in a not ready state

charts/seed-monitoring/charts/prometheus/rules-tests/kube-scheduler.rules.test.yaml

@@ -6,6 +6,7 @@ evaluation_interval: 30s
 tests:
 - interval: 30s
   input_series:
+  # KubeSchedulerDown
   - series: 'up{job="kube-scheduler"}'
     values: '0+0x30'
   alert_rule_test:

charts/seed-monitoring/charts/prometheus/rules-tests/kube-state-metrics.rules.test.yaml

@@ -5,10 +5,13 @@ evaluation_interval: 30s
 tests:
 - interval: 30s
   input_series:
+  # KubeStateMetricsShootDown
   - series: 'up{job="kube-state-metrics", type="shoot"}'
     values: '0+0x20'
+  # KubeStateMetricsSeedDown
   - series: 'up{job="kube-state-metrics", type="seed"}'
     values: '0+0x20'
+  # NoWorkerNodes
   - series: 'kube_node_spec_unschedulable'
     values: '2+0x20'
   - series: 'kube_node_spec_unschedulable'

charts/seed-monitoring/charts/prometheus/rules-tests/machine-controller-manager.rules.test.yaml

@@ -6,6 +6,7 @@ evaluation_interval: 30s
 tests:
 - interval: 30s
   input_series:
+  # MachineControllerManagerDown
   - series: 'up{job="machine-controller-manager"}'
     values: '0+0x30'
   alert_rule_test: